Understanding the Relationship between Urban Form and Urban Shrinkage among Medium-Sized Cities in Poland and Its Implications for Sustainability

Abstract

Understanding the relationship between urban form and urban shrinkage is crucial for developing sustainable urban policies, particularly in medium-sized cities facing demographic and economic challenges. This study investigates the complex relationship between urban form and urban shrinkage in medium-sized Polish cities (population of 20,000 to 100,000), highlighting the implications for sustainability. Utilising a comprehensive multi-factor approach, it analyses the shrinkage and growth trends over 15 years (2006–2021) by establishing a shrinkage/growth score based on social, demographic, and economic factors for each city. It examines spatial aspects, particularly urban form compactness and population density, using Corine Land Cover (CLC) spatial data, making the methodology applicable to urban areas across Europe. The findings reveal no significant overall correlation between urban compactness and shrinkage/growth score across all cities. However, a positive correlation exists within “urban municipalities”, indicating that less compact urban areas tend to experience more shrinkage. Additionally, a temporary negative correlation between population density and shrinkage/growth score was observed from 2006 to 2016, which shifted to a positive trend in “urban municipalities” from 2016 to 2021. These results highlight urban shrinkage’s complex and dynamic nature and its potential ties to urban form. The study concludes with recommendations for urban policymakers and planners regarding compact and dense urban strategies to mitigate the adverse effects of shrinkage and enhance urban resilience and sustainability. While the trends change, the study highlights the need for further analysis of these relationships.

1. Introduction

The modern-day discussion on urban population decline began in the second half of the 20th century when so-called “urban shrinkage” became a common transformation for many large cities in Europe and North America. German researchers Hausermann and Siebel [1] first used the term “shrinking cities” to describe long-term demographic, socio-economic, and spatial changes in urban areas. Since then, research on urban shrinkage has been conducted around the globe. According to the most recent UN-Habitat—World Cities Report 2022, almost half of the developed countries’ cities have experienced shrinkage processes since the 2000s, most based in Europe and North America [2]. Significant population losses have also been recorded in Japanese and Northeastern Chinese cities [3,4], as well as in Australian small towns. Shrinkage is always linked to specific local and global forces. In Australia, it is linked to youth migration, fluctuation in global mineral markets, climate change, and policy change in the localisation of government services [5,6], while urban shrinkage in the Global South is additionally related to conflicts often exacerbated by climate change [2]. Because of its multidisciplinary nature and complexity, understanding this phenomenon still challenges researchers worldwide.

The first definition of urban shrinkage (Schumpfungsprozess) embraced the multidimensional aspects beyond the population decline problem and summarised it as follows: “… [it] does not lie in individual developments. Only the interaction of population losses with selective migration of qualified young workers, unsuccessful integration of immigrants, negative economic developments, high unemployment, declining municipal financial leeway, dissolution of the city structure, and thinning out of the supply of goods and services results in an urban crisis in which negative developments can intensify into a vicious circle” [1] (p. 10). In some shrinking cities, adverse developments are deteriorating the fiscal base and disturbing the maintenance of local infrastructure levels and quality of life. Consequently, many are suffering from problems related to vacant and underutilised housing, uncompetitive local business firms, and derelict transportation systems, including streets and other utility infrastructure [7]. Large amounts of housing vacancies or underutilised supply networks and facilities raise the question of whether such infrastructures can be sustained [8]. When prolonged, this state can lead to a decline spiral, deepening over time [9].

In general, it was observed that smaller cities near economically prosperous metropolitan areas benefit from urbanisation effects and can compensate for natural population development. In contrast, those in peripheral and structurally weak areas struggle to cope with the consequences of demographic change [10]. Due to limited resources, small- and medium-sized cities with less beneficial positions will have less time to react. The question of whether effective urban spatial strategies on a local level can influence resilience to the decline spiral and promote a sustainable path for shrinking medium-sized cities without growth is an important and complex one. Resilience theories suggest that cities can adapt, transform, or resist challenges and maintain their core functions [11]. However, it is essential to understand which spatial qualities impact urban resilience and sustainability. Reis, Silva, and Pinho [12] note the growing importance of understanding the relationship between spatial qualities and urban resilience to urban shrinkage. However, understanding how effective urban strategies can influence resilience and promote sustainability in shrinking cities requires a comprehensive and interdisciplinary approach.

Urban planning practices and how they relate resilience to urban decline have been widely debated among scholars [9,13,14]. According to Schwarz et al. [15], shrinkage has severe implications for all dimensions of sustainability and quality of life in cities. These implications force planners and policymakers to search for new concepts. Pallagst et al. [9] suggest that effective management of urban shrinkage requires shifting attitudes from growth to de-growth, reuse, and retrofit. In times of scarce external funding, inclusive governance and innovation in urban planning play a crucial role [9]. Haase et al. [14] point out that different external and internal drivers influence urban growth and development, and planning policies should focus on inward-oriented, right-sized, compact cities with revitalized dense urban centres. The Urban Renewal program in Eastern Germany successfully implemented this approach through the Stadtumbau Ost program and the complementary IBA Saxony–Anhalt 2010 program for medium-sized cities facing urban shrinkage. These projects provided valuable insights for research and planning for urban shrinkage. Despite criticism, they enabled many cities in Eastern German states to enter a new path of stable de-growth despite continuous depopulation trends in the region. The abovementioned approach emphasises urban compactness and density as the critical qualities that help redirect the decline path towards a more optimistic, stable de-growth path.

The concept of urban compactness has been widely linked with the sustainability of urban areas. In the context of shrinkage, it has been associated with improved outcomes of sustainable shrinkage planning. However, the empirical literature on this subject is still lacking. A literature review conducted by Reis et al. [12] on spatial metrics used in urban shrinkage studies revealed that these metrics are insufficient to comprehensively assess spatial patterns.

In light of this, the present study aims to at least partly bridge this gap in empirical literature by testing the relationship between urban compactness, urban population density, and urban shrinkage processes. The objective is to assist urban governments in responding more precisely to the challenges of urban shrinkage in medium-sized cities by providing them with a methodology to monitor the urban shrinkage patterns with the use of available datasets. While more research on a wide, national scale is still needed to gain understanding, this study serves as a step in finding the relationship between urban form and urban shrinkage patterns in European medium-sized cities.

The conceptual framework for the study is based on the example of Poland. Possible connections between changes in urban form in all Polish medium-sized cities (population size of 20,000–100,000) and urban growth and shrinkage processes are investigated. These proposed connections reflect current planning knowledge regarding factors associated with urban shrinkage. Against the background of the identified research gaps, briefly summarised in the introduction and expanded in subsequent sections, together with reflection on the debate and theory of shrinking cities, the following hypotheses structure this work:

H1. 

There is a statistically significant correlation between compactness and shrinkage of medium-sized Polish cities.

H2. 

There is a statistically significant correlation between urban population density and shrinkage of medium-sized Polish cities.

H3. 

The trend persists within the analysed timeframe.

This study incorporates statistical and geospatial data to explore urban shrinkage and growth processes and the shape of each city in the analysed years. Geospatial tools evaluate urban form concerning its compactness and population density. By correlating the findings of both datasets, this study aims to test the hypothesis mentioned above. Figure 1 provides an overview of the study’s structure and methodology. Within this diagram, a number of key tools have been used to execute the research. Briefly, these are as follows: QGIS for spatial analysis and Visual Studio Code to manage and write scripts in Python language with Pandas libraries.

The article has six sections. After this general introduction, Section 2 addresses the changing spatial patterns of urban shrinkage, presenting an overview of the literature. A particular emphasis is set on the measures of urban form. This section concludes with identified research gaps and outlines the research proposal. Section 3 presents the materials and methods used in the research, along with the selected subjects and time frames of analysis. The outcomes of all measured variables are presented in Section 4, along with a statistical analysis of their relationships. Finally, Section 5 presents and compares the main findings with previous research. The article ends with Section 6., which summarises the conclusions and provides a list of references.

2. Spatial Patterns of Urban Shrinkage

Although analysing urban morphology and shape in urban research dates back to at least the early 20th century [16,17,18], data and computational limitations have prevented a more comprehensive implementation of these metrics beyond case studies. However, the field has significantly broadened with the advancement of geospatial tools and computing power. Nonetheless, little research has examined urban metrics and shrinkage across all cities in the national urban network, which necessitates significant computing capacities and sophisticated spatial data analysis. Recent research in this field highlights the importance of quantitative methods and emphasises the physical dimension of urban areas.

2.1. State of Research on Spatial Patterns of Urban Shrinkage

According to Kazimierczak and Szafrańska [19], there are three primary scales for analysing spatial patterns of urban shrinkage, as shown in Figure 2. Comparative studies at the national and regional levels can identify differences and similarities in the causes, progression, and consequences of depopulation among a specific group of cities, providing general and detailed conceptual and theoretical assumptions (e.g., Turok and Mykhnenko [20]). At a lower level, general urban studies (urban and supra-urban scale, according to Kazimierczak and Szafrańska [19]) can identify direct and indirect depopulation effects relevant to local growth and planning. Intra-urban scale studies, which focus on a city, its specific districts, urban zones, or housing estates, can be compared across the city to provide a comprehensive understanding of the dimensions of urban shrinkage [19]. It can also reveal shrinkage of certain areas despite overall growth processes when looking at the entire city. Thus, integrating studies at all three levels can provide a complete picture of urban shrinkage.

A literature review on spatial patterns of urban shrinkage conducted by Reis, Silva, and Pinho [12] shows that it is less extensive than the literature on growth patterns. This is partially because spatial patterns of shrinkage tend to be less precise, as built-up areas do not disappear when people move out [12]. Großmann et al. [21] propose examining shared spatial characteristics across different scales that may indicate shrinking urban areas.

2.1.1. Regional or Inter-Urban Scale

Spatial patterns of urban shrinkage on a national scale often reveal entire regions undergoing structural changes such as deindustrialisation. Examples can be found in the Rust Belt in the USA, the old industrial regions of Northern Great Britain, the German Ruhr region [9], China’s Northeastern regions of the Yangtze River Delta [22,23], and many more worldwide. These hot spots of urban shrinkage might be temporal or long term. Analysing large-scale associations involves looking at metropolisation processes and their effects on all cities in the urban hierarchy. Liu et al. [22] looked at inter-urban hierarchical dependencies to analyse patterns between different sizes of growing and shrinking cities in Northeast China from 2000 to 2020. Their study reveals five typical patterns, with small and medium towns experiencing more challenges than bigger, central ones. However, the growth of the central city might as well have a strong spillover effect on the region [22]. Tan et al. [23] used Nighttime Light Data to identify possible reasons for urban shrinkage in the Yangtze River Delta. The remotely sensed nighttime light (NTL) can accurately record the trajectory of human production and life, reflecting the integrated demographic, economic, and spatial changes of cities. Their study concluded that the mono-economic structure, the difficulty of industrial transformation, and the lack of linkage among bigger cities were the leading causes of shrinkage in that region [23]. Yu et al. [24] investigated the connection between urban shrinkage and urban resilience indicators. Their research pointed to various resilience indicators, including urbanisation and environmental indicators. The results showed no significant relation between shrinkage and these indicators. However, the nature of Chinese urbanisation and its hierarchical urban settlement model makes it difficult to relate the results to other urban contexts. Therefore, research conducted in the European context can show different relationships compared to those found in China.

2.1.2. Urban and Supra-Urban Scale

When analysing urban areas in Germany in the period from 1996–2006, Siedentop and Fina [25] observed increased suburban sprawl, which continued despite ongoing depopulation in the core city. The physical patterns of “shrinkage-sprawl” are similar to those of urban sprawl in a growing context, resulting in a fragmented and perforated territory with low-density development, increasing vacancy, and deteriorating urban fabric in inner city locations (Figure 3). On the other hand, research on “post-socialist sprawl” conducted by Schmidt et al. [26] demonstrates that declining densities and sprawling growth patterns are not only linked to demographic decline and economic change, but instead, they have been “exacerbated by public policy and unregulated market-induced growth in the case of the other CEE countries” [26] (p. 17). They conclude that the continuing demographic transition of much of Eastern Europe calls for a research agenda that analyses the impacts of shrinkage on urban development in much more detail [26].

Urban shrinkage often results in urban areas where fewer people and fewer activities are spread out across a more extensive territory [25,27]. This results, on the one hand, from low-density suburban sprawl and new land-intensive industrial enterprises, an automobile-oriented transport network, and, on the other hand, a deteriorating urban core that suffers from population loss, a lack of investment in public infrastructure, neighbourhood deterioration, and an overall image problem [26]. Consequently, this development pattern generates excessive costs, such as increased public expenditures for building and maintaining infrastructure and public services. It bears a commercially negative impact on the city centre, an increase in energy and fuel consumption, and a negative impact on household budgets and the environment [28]. In a study on the correlation between urban sprawl and economic performance in selected growing Polish cities, Lityński [29] concluded that higher local economies significantly correlate with a more minor degree of urban sprawl.

Consequently, the more compactly the houses in a particular municipality are built across its space, the higher the local economy level, regardless of the distance from the city [29]. On the other hand, a study of urban sprawl in two shrinking cities, Leipzig and Liverpool, shows that sprawl is not a challenge bounded by shrinkage [27]. The study concluded that declining cities’ suburban zones are similar to those in thriving regions. However, some shrinking cities might have problems with urban sprawl, but they are often more burdened with the urban decline of the inner city [30].

2.1.3. Intra-Urban Scale

The growth and shrinkage of urban areas often co-occur, resulting in a growing geographic divide within and between cities and their surrounding regions. Various studies, including those by Hollander et al. [7], Pallagst et al. [3], and Wolff and Wiechmann [31], have examined this phenomenon. A study by Schwarz et al. [15] tested simulation models on urban land-use change to understand relations with shrinkage. The models were tested in selected US, European, and Japanese cities. The results show a very heterogeneous mechanism; no single model fulfils the criteria to explain the relations. Another study explored the morphological aspects of Łódź and found that patterns of intra-urban shrinkage were not uniform, with the downtown and historic areas being particularly affected, in line with Klaassen’s [32] theoretical assumptions regarding the mismatch between urban and social subsystems [19]. However, while there are numerous case studies on the subject, large-scale and longitudinal studies on the spatial patterns of urban shrinkage are few. One such study by Wang et al. [4] analysed 15 cities in Northeastern China and found that shrinking cities had lower compactness and land-use efficiency than growing ones. As cities evolve and transform, industries that rely on resources may abandon significant amounts of industrial land, exacerbating the problem of lagged land-use efficiency in shrinking cities. Further research on this topic is necessary to identify spatial patterns of urban shrinkage in Europe and provide more evidence for informed planning decisions.

2.2. Polish Urban Shrinkage Context

The decline in the urban population in Poland is a significant challenge for the country’s spatial policy, as projected by the Poland Statistics [33] forecasts for 2023–2060. According to the report, the population of Poland will experience a decline by 2060 of 25% in the low scenario and 40% in the high scenario [34]. This decline will primarily affect urban areas as the progress of suburbanisation is expected, especially in rural areas around major urban centres [34]. This alarming trend has prompted numerous researchers to address the issue and propose strategies to counteract and eliminate its adverse effects in cities. The Institute of Geography and Spatial Organization of the Polish Academy of Sciences has conducted numerous analyses of this topic. Project manager prof. Śleszyński stated that, proportionally, it might be one of the most significant declines in the world.

Furthermore, medium-sized cities are the group most affected [35,36], and due to limited resources, they face challenges to manage it quickly and effectively [35,36,37]. A study by Szymczyk and Bukowski [38] confirmed that trend by investigating the patterns of urban shrinkage in all Polish cities through a multicriteria indicator approach. It highlights that medium-sized cities were the predominant size type among all shrinking cities between 2016 and 2021, with over half of medium-sized cities classified as shrinking, as shown in Figure 4. It is worth adding that the number of inhabitants of shrinking medium-sized cities (4.0 mil. in.) is almost the size of small and large shrinking cities together (4.7 mil. in.). The terms small, medium, and large cities in the Polish context are explained later in this paper.

In the European settlement network, medium-sized cities have a vital supply function for rural regions, providing goods and infrastructure for basic and more specified needs. They are essential elements of a sustainable polycentric urban network. These types of cities are especially endangered by demographic change and fiscal collapse. Their future significance and development depend on external circumstances and internal governance strategies. Due to the urgency and magnitude of the problem, the urban shrinkage of medium-sized cities in Poland challenges traditional, growth-oriented urban planning strategies and requires informed, innovative approaches. Thus, empirical studies on the relationship between shrinkage and urban space help find more adequate solutions.

While medium cities are the group most affected by shrinkage in Poland, many are in socio-economic balance or thriving. Understanding if there are common spatial patterns among growing and shrinking medium-sized cities can provide critical information about relations between spatial and socioeconomic factors. Since planning theories and tools were mainly developed for growth-oriented planning, researchers call for advancing shrinkage-oriented tools [12] to help shape more appropriate urban shrinkage governance.

2.3. Compactness

Urban compactness is typically viewed as the antithesis of urban sprawl [40], linked to high social and environmental costs in urban planning [41]. While the term “compact city” has different meanings, it implies different evaluation approaches. The OECD describes the characteristics of the compact city as “dense and proximate development patterns (…) urban areas linked by public transport systems (…) accessibility to local services and jobs” [42] (p. 15). The term “compact city” is often said to have first been used by Dantzig and Saaty [43], who were principally interested in a more efficient use of urban resources. According to Ahlfeldt and Pietrostefani [44], three concepts are commonly used in describing the compact city: economic, morphological, and mixed-use densities, each translating into different measures (

Table 1. Compact cities characteristics. Source: Ahlfeldt and Pietrostefani [44].

Index	Characteristic	Summary
A	Economic density	Refers to the number of economic agents living or working within a spatial unit and is typically measured as population or employment density (Thomas and Cousins 1996; Churchman 1999; Burton 2002; Neuman 2005).
B	Morphological density	Refers to the density of the built environment and captures aspects of the compact city such as compact urban land cover, demarcated limits (demarcated urban/rural land borders), street connectivity, impervious surface coverage, and a high building footprint to parcel size ratio (OECD 2012; Wolsink 2016; Neuman 2005; Burton 2002; Churchman 1999).
C	Mixed land use	Captures the co-location of employment, residential, retail and leisure opportunities (Churchman 1999; Burton 2002; Neuman 2005), both horizontally across buildings and vertically within buildings Burton (2002).

). They include measures such as the population density of a spatial unit, employment density, and mix of uses or demarcated limits of urban and rural borders [44]. It is worth adding that compactness is regarded as a relative concept, not a concept of absolute material density. Thus, it must be analysed in a specific national or regional context.

An example of such a study can be found in Angel et al. [45], where the authors explore relationships between the compactness of urban form and climate change. They state: “Other things being equal, both population density and shape compactness help determine the average travel distances in cities, and, hence, affect their energy consumption and their greenhouse gas emissions. They also affect the length of infrastructure lines and the length of commutes. In principle, therefore, increasing either the shape compactness or the population density of cities can contribute—in different yet similar measures—to mitigating climate change” [45] (p. 1). Overall, the study not only demonstrates the importance of density and shape compactness as critical measures in the shrinkage discussion but also as critical in helping to forge greater sustainability, especially in urban transport and climate change policy in cities.

Another example of measuring compactness comes from China. In their research on spatial relations of shrinking cities, Wang et al. [4] used a multicriteria assessment with multiple statistical indicators ranging from population, economic, land use, and public-service compactness [4]. Mouratidis [46], on the other hand, uses a combination of urban population density, building typologies, and functional mix in his study to assess the compactness of Oslo’s neighbourhoods. In a study on the link between urban form and CO₂ emissions, Guo et al. [47] used four measures to analyse urban area compactness and complexity. He et al. [48] analysed 293 Chinese shrinking cities in search of relationships between shrinkage and urban form characteristics. Their research examined changes in morphological characteristics of urban expansion, such as fragmentation, compactness, urban sprawl, and city size.

Due to the scale of this study (a large number of analysed data), it uses selected key indicators and follows the “morphological density” characteristics Ahlfeldt and Pietrostefani [44] described. It aims to analyse the spatial urban form of all medium-sized municipalities in Poland, reflected by the urban–rural boundaries and measured with “shape irregularity” [12] (p. 11). The authors calculated compactness using a compactness index of demarcated urban borders and more precise urban population density, considering the actual urbanised land instead of administrative boundaries. Such politically drawn boundaries can arbitrarily include large areas of non-urbanised territory or, conversely, very little or no non-urban land, depending on where they are drawn. Densities so derived are, therefore, totally inconsistent with each other and non-comparable.

3. Materials and Methods

This study adopts a quantitative approach to examine the urban settlement network in Poland over a 15-year timeframe. It draws on classifying urban areas in Poland using a multicriteria indicator of urban growth and shrinkage. The results from the multicriteria analysis are then compared to spatial data expressed in compactness and density. The study checks if there is a correlation between urban growth and shrinkage and urban-form parameters. The following section provides a comprehensive overview of the study’s subjects, time frame, data, and analysis methods.

3.1. Subjects

This study focuses on all medium-sized Polish cities. The selection of cities is based on Poland’s administrative system. Expressed simply, on the highest/regional level, we have a voivodeship (województwo), which consists of counties (powiat); counties are divided into individual local-authority municipalities (gmina); a municipality contains either an individual city (GUS level 6, kind 1), only villages—rural area (GUS level 6, Kind 2), or a mix of a town and villages—urban-rural area (GUS level 6, Kind 3)(see Figure 5 for municipalities map). Due to data availability, continuity of research, and nationwide comparability, the selected method uses the smallest local government unit, i.e., the municipality (gmina).

In 2021, Poland had 2477 municipalities: 302 urban, 662 urban–rural, and 1513 rural. While this research focuses on medium-sized cities and towns (in Poland, both cities and towns are called “miasto”, a city), we utilised data for “urban” municipalities and those in the “urban–rural” category. An “urban municipality” (6, 1) is a municipality with an individual city and has administrative boundaries tightly enclosing urban areas. An “urban–rural municipality” (6, 3), on the other hand, includes a city and various urban areas such as villages and small towns loosely connected to each other. Due to the difference between these two types of urban areas, the analysis is divided into two categories:

-

urban area in the urban municipality (6, 1),

-

an urban area in the urban-rural municipality (6, 3), including a city and small settlements.

To maintain data consistency, we carefully considered the varying number of spatial units in Poland over different years. After much consideration, we ultimately selected 964 cities to ensure accurate comparison and analysis of our data without discrepancies. We relied on a simplified division based on Statistics Poland’s categorisation. Cities with a population of up to 20,000 are categorised as small, between 20,000 and 100,000 as medium, and above 100,000 as large. When analysing urban population data, it is crucial to establish a consistent reference point for size classification. Therefore, we chose to base our categories on the beginning of each analysed period despite potential population fluctuations over time. This approach allowed us to guarantee precision and consistency in our analysis. The population was categorised based on the 1st year for the 5-year periods. Figure 5 provides a map of the different spatial entities.

3.2. Timeframe

Urban dynamics are inherently complex and can vary depending on the specific phenomena being studied. For instance, socio-economic factors tend to be more dynamic than physical changes to the built environment. To fully understand the relationships between these factors, it is necessary to conduct longitudinal studies. The timeframe of our study was chosen for two reasons: first, because of the availability of spatial data (as discussed in Section 3.4), and second, because our study builds on the multicriteria indicator assessment of shrinking cities in Poland conducted by Szymczyk and Bukowski [38]. The analysis was divided into three sub-periods: 2006–2011, 2011–2016, and 2016–2021 (Figure 6).

3.3. Statistical Data

The literature aiming to quantify the phenomenon of urban shrinkage can be divided into two fundamental approaches [50]. The traditional view sees urban shrinkage as a progressive depopulation, with studies primarily focusing on population-based indicators [20,51,52]. However, there are examples where demographic and economic development do not align. Cities with declining populations may still have robust economic structures and continue to thrive [31]. In Poland, a multi-criterion approach to assess urban shrinkage has been conducted by researchers such as Jaroszewska [53], Śleszyński [36], Sroka [54], and Szymczyk and Bukowski [38], among others. Of these, Szymczyk and Bukowski [38] focused on the method developed by Milbert [55] from the Federal Office for Building and Regional Planning in Germany (BBSR in German). The Milbert method captures this multidimensionality through six demographic and socio-economic indicators. This method was adapted to the Polish territorial division and statistical data. It considered the following equivalent indicators from Polish Statistics: annual average population development (in %), net migration per 1,000 inhabitants, annual average change in working-age population (in %), annual average change of employed persons (in %), annual average change of unemployed persons in (% points), and annual average change in municipality own income per inhabitant (in %). A summary of the indicators with different units of measurement and a spread of the division into quintiles is made. Based on a tested normal distribution, the limits of the outer quintiles are always such that the relative distance from extreme values and outliers does not play a role. The affiliation of units to a quintile is translated into points for each indicator. Points are summed up into a numeric score ranging from 0–24 for each city in each research period. Cities scoring 0–10 are considered shrinking, 11–13 are stagnating, and 14–24 are growing (the calculation outcomes can be found here: https://rcin.org.pl/dlibra/publication/276466 (accessed on 13 May 2024). The outcomes for a group of medium-sized cities are selected to relate them with the spatial analysis results we conduct here. Figure 7 shows the distribution of different types of medium-sized cities in 2016–2021, with division from strongly growing to shrinking significantly. Consequently, we have analysed all 180 medium-sized cities in the first period, 184 cities in the second period, and 180 cities in the last period.

3.4. Spatial Data

The data for intra-urban spatial metrics of medium-sized cities is based on land-use data from the CORINE Land Cover (later called CLC) for the area of Poland. The CORINE Land Cover (CLC) data is one of the most important sources of land use and landscape dynamics data from a European perspective and the only consistent and harmonised multinational, continental-scale collection of land-use information in the world [56]. It is a standardised methodology for producing continent-scale land cover, biotope, and air-quality maps. CLC uses a hierarchical classification system to categorise land cover and land use into 44 classes grouped into the following five main categories:

• Artificial surfaces—built-up areas, including residential areas, commercial and industrial areas, mines, and green urban spaces.
• Agricultural areas—arable land, permanent crops, meadows, pastures, and land principally occupied by agriculture with significant areas of natural vegetation.
• Forests and semi-natural areas—forests, shrubs, and open areas with little or no vegetation.
• Wetlands—inland marshes, peat bogs, salt marshes, salines, and intertidal flats.
• Water bodies—inland waters and marine waters.

The methodology involves the interpretation of satellite images combined with ancillary data and field surveys to ensure accuracy (see a detailed explanation of the CLC methodology in the European Environment Agency [57]). Furthermore, CLC classes can be associated with the concept of defining Production–Living–Ecological Space (PLES), which is commonly used in Chinese planning policies. PLES categorises land into spaces for production (industrial and commercial units in CLC), living (urban fabric in CLC), and ecological purposes (forest, wetlands, and water bodies in CLC). This alignment allows for the integration of PLES into the European context. Since the product is updated every 6 years, with the most recent update in 2018, the three corresponding datasets were selected for this analysis. For 2006–2011, a CLC2006; for 2011–2016, a CLC2012; and for 2016–2021, a CLC2018 (Figure 6). In their study of urban sprawl in Poland, Cieslak et al. [58] indicated that the CLC database is a reliable source of information about urbanisation processes. On the other hand, it has to be noted that there are some discrepancies between the CLC and the actual urbanised areas in Poland [59]. Consequently, the results were later compared with the data on urban population density calculated by the Head Office of Geodesy and Cartography (GUGiK) in Poland [60], which employs similar urban uses (see

Table 2. CLC codes indicate classes considered and their respective colours. Source: CLC [56].

	CLC Code	Name
	111	Continuous urban fabric
	112	Discontinuous urban fabric
	121	Industrial or commercial units
	122	Road and rail networks and associated land
	123	Port areas
	124	Airports
	131	Mineral-extraction sites
	132	Dump sites
	133	Construction sites
	141	Green urban areas
	142	Sport and leisure facilities

) but a different calculation method.

3.5. Defining Urban Compactness

To define the shape of the urban area, we considered the 44 classes of land uses provided by the CLC. We opted for two approaches. First, we consider an urban area a shape represented by only “urban fabric” layers (Class Number 111, 112). This is where most people live in the city. Secondly, we also analyse a more broadly defined urban area as “artificial surfaces”, which includes infrastructure and green urban areas, represented by 11 classes (Table 2). The aim is to ensure that the two approaches allow us to control for potential errors related to inadequate urban form definition.

Built-up and urbanised areas, according to GUGiK, are shown below in

Table 3. Polish Statistics codes and classes indicating urbanised areas (in Polish, “grunty zabudowane i zurbanizowane”). Source: GUGiK [60].

Code	Class Name
B	Housing
Ba	Industrial
Bi	Other built-up areas
Bp	Urbanised areas with no buildings
Bz	Recreation areas
K	Mining areas
dr	Transport areas—Roads
Tk	Transport areas—Rail
Ti	Transport areas—other
Top	Areas designated for future infrastructure investments

.

The selected classes were isolated and cut to the boundaries of the municipality. Next, they were merged into one shape consisting of one or more polygons (Figure 8). Each shape received a unique identification number of a spatial unit, called the National Official Register of the Territorial Division of the Country (TERYT).TERYT functions on the basis of The Law of June 29th, 1995 on official statistics (Journal of Laws from 2018, item 997 with amendments) and the Regulation of the Council of Ministers of December 15th, 1998 on detailed principles of running, using and making available the territorial register and related to it, obligations towards the state administration agencies and the self-government entities (Journal of Laws from 1998 No 157, item 1031 with amendments). Moreover, the development of the analytical method included some calibrations. Due to some discrepancies in the spatial data (such as imprecise land use demarcation), many cities had minor shapes left after cutting to the boundaries. We have analysed possible thresholds to avoid distorting the calculation by examining the histogram of shape sizes. Most small patches had a size below the 2% of the total shape area. Consequently, we have decided to set a threshold of 2% of the total shape area. All the shapes that were below this value were deleted.

In summary, the calculation consists of the following operations (presented in Figure 8) for each selected year:

• Isolation of selected land use classes;
• A cut of land-use polygons with city administrative boundaries;
• Merge polygons into continuous shapes;
• Delete polygons below a set threshold for a particular shape;
• Calculate the shape’s area and perimeter;
• Calculate the compactness index;
• Calculate the urban population density.

3.5.1. Spatial Measures

Once the shape is extracted and its measures calculated, we calculate a measure of compactness and a measure of urban population density. For this research, the following definitions are used:

AD	Urban area
AP	Urban area perimeter length
Ci	Compactness index
Pop	Urban population
Pd	Urban population density

Compactness Index Calculation

As mentioned previously, there are multiple ways to measure compactness. This study uses the compactness index (Ci) measure, quantifying the compactness of shapes’ irregularity. According to the classification by Reis et al. [12], the proposed metrics “measure the extent to which urban settlements are more continuous and concentrated or more scattered (fragmented)” [12] (p. 13). Thus, the fundamental idea primarily revolves around the two-dimensional expansion pattern of an urbanised area, which is deemed to be more compact if the pattern is more clustered towards a centre and with less sprawl, leapfrogging, or branching [61]. There are many measures of the compactness index [62,63,64]. Among them, a few were identified by Barnes and Solomon [65] as the most commonly used. Among these most common, the authors selected a single measure named after the name of its author, Schwartzberg [17].

Interestingly, both the Polsby–Popper and Schwartzberg Compactness indexes are mathematically equal (one score is the other score raised to a power) and based on comparing an urban area shape to a circle assumed to be a perfectly compact shape. Schwartzberg compactness index is calculated as the ratio of the perimeter of the urban area (A_D) to the circumference of a circle whose area is equal to the area of the urban area. It can be written as

$$Ci={\displaystyle \frac{2\pi \sqrt{A_D/\pi }}{A_P}}$$

Scores range from 0 to 1, where 0 is the least compact and 1 is the most compact. This index was calculated for each medium-sized city in each CLC year.

Urban Population Density Calculation

The second measure considers the urban area (A_D) and the municipality population (Pop) in the corresponding period to calculate the urban population density. As opposed to the city population density provided in the national statistics, this method captures a more accurate density as it considers not the administrative boundaries but the actual urban area within them. Compared to the administrative boundaries, the urban area excludes agricultural land, forests, and water bodies, among others, providing a more accurate picture of the city boundaries and where people live. The urban population density (Pd) can be formally written as:

$$Pd={\displaystyle \frac{Pop}{A_D}}$$

3.5.2. Statistical Correlation Analysis between Spatial Measures and Urban Shrinkage

In the last step, an analysis was undertaken, establishing the correlations between the urban growth and shrinkage score, the urban-shape-compactness measure (Ci), and the urban population-density measure (Pd). We conducted the correlational analysis, determining whether there is any relationship or association between several variables of interest [66]. A correlation analysis and Pearson’s correlation coefficient (r) were used. Correlation r results from comparing two variables in their normalised form over the sample of n values in each case. One of its properties is that its value varies between +1 and −1, representing a maximum positive and inverse correlation. Values close to 0 show there is no statistical correlation between the variables. In the social sciences, values above 0.5 can be considered a statistically significant, high correlational effect [67]. The analysis method used the statistical significance index or p-value. This indicator confirms that the correlation between two variables is significant if p is below 0.05. This enabled us to test our research hypotheses. It is worth mentioning that correlation results are cross-sectional and should be interpreted cautiously, as they point to associations but not necessarily causal relationships.

3.6. Tools

Due to the spatial scope, long timeframe, and volume of data (multiple variables, times all 2621 municipalities considered in this research, times three periods), any manipulation required data-analytic tools. Python 3.11 programming language and data analytics libraries (Pandas and NumPy) were used to harmonise and clean the data. After downloading, raw CLC data were converted into a dataset connecting the statistical ID with spatial ID (called TERYT). Each spatial unit was checked for completeness. Spatial analysis was conducted using QGIS 3.28 and Python scripts (see the Appendix A for detailed information).

4. Results

The results are divided into two stages. The first stage (Section 4.1) shows the outcomes of each variable analysis: compactness index (Ci) and urban population density (Pd). The second stage (Section 4.2) shows correlation analyses between the variables and urban growth and shrinkage score (Sc) (for more details on Sc, see Section 3.3).

4.1. Spatial Measures Calculation Outcomes

This section presents the outcomes of calculating compactness index (Ci) and urban population density (Pd) for three selected CLC datasets (2006, 2012, and 2018).

4.1.1. Compactness Index

The results of the compactness analysis for medium-sized cities in three periods show rather similar outcomes (

Table 4. Outcomes of the compactness index analysis (Schwartzberg compactness index) for all medium-sized Polish cities in 2006, 2012, and 2018 considering urban areas as “artificial surfaces”.

Schwartzberg Compactness Index Outcomes Per Year	2006	2012	2018
Number of observations	180	184	180
Min	0.1160	0.1047	0.1049
Median	0.2628	0.2577	0.2503
Max	0.5146	0.5221	0.5186

). The Schwartzberg compactness index is expressed in numbers ranging from 0–1, with one being the most compact. It shows a normal distribution in all years. While the maximum values decreased in time, the minimum ones in 2006 increased slightly in 2012 and 2018. This shows an overall trend of light dispersion in small- and medium-sized cities, while large cities did not change their compactness index over time). Among the most compact medium-sized cities throughout the analysed timeframes (Holding top 10 position in all the analysed periods) were Ząbki, Piastów, Żyrardów, Legionowo, Słupsk, Giżycko, Świdnica, and Rumia. The least compact were Wyszków, Świebodzin, Opoczno, Jastrzębie-Zdrój, and Pszczyna (Figure 9).

4.1.2. Urban Population Density

The urban population density-analysis results for medium-sized cities in three periods show a normal distribution and a strongly decreasing trend for all city sizes (

Table 5. Urban population density for medium-sized Polish cities in 2006, 2012, and 2018 considering urban areas as “artificial surfaces”.

Urban Population Density Outcomes Per Year	2006	2012	2018
Number of observations	180	184	180
Min (ppl/km²)	1511	1512	1406
Median (ppl/km²)	3259	3000	2903
Max (ppl/km²)	5353	4893	4502

). While the maximum values decreased in time, the minimum densities fluctuated, showing a trend of medium cities becoming less dense over time. This shows an overall trend of dispersion in medium-sized cities. Among the densest cities throughout the analysed timeframes (Holding top 10 position in all analysed periods) were Wejherowo, Sopot, Świętochłowice, Ełk, and Grudziądz (Figure 9). Polkowice, Trzebinia, Skawina, Czechowice–Dziedzice, Wyszków, and Pszczyna were the least dense.

It is important to note a discrepancy between the results obtained from the 2012 CLC dataset calculation and the urban population density available in Polish Statistics (

Table 6. A comparison of selected cities’ urban population density based on CLC and GUGiK data. Source: Author’s research and GUS BDL [33].

Name of the City	Urban Population Density Based on the Authors’ Calculation for ‘Artificial Surfaces’ for CLC2012 (people/km²)	Urban Population Density Sourced from GUS BDL for 2013 (people/km²)
Wieliczka (6, 3)	2134	3634
Jasło (6, 1)	2357	3568
Bartoszyce (6, 1)	3809	4295
Śrem (6, 3)	3156	2877
Leszno (6, 1)	3307	4164
Wejherowo (6, 1)	5735	5847
Sopot (6, 1)	5614	5582
Trzebinia (6, 3)	1617	2926
Polkowice (6, 3)	1513	1231

). However, the Polish Statistics data were only published for 2013, 2014, and 2015, with no new data available to date, making it difficult to compare. Based on the analysed sample of randomly selected cities, the Polish statistics appear to calculate fewer urban land areas than the “artificial surfaces” do (and more than “urban fabric”). As a result, density values are higher in Polish Statistics. There could be several reasons for this difference. While Table 2 and Table 3 show that the principles of delimitation of urban areas are very similar, the method relying on CLC uses satellite images, which simplifies the contours of urbanised land, which leads to discrepancies. On the other hand, CLC 2018 dataset classes were compared with building address points data by Śleszyński, Gibas, and Sudra [59], showing an average of approximately 35% of buildings located in areas defined in CLC as agricultural, suggesting an underestimation of urbanised areas in CLC. Finally, since this study aims to establish a methodology that uses comparable, open-source, up-to-date, and Europe-wide land-use data, CLC data meet these requirements. While Polish Statistic data are not regularly updated, it does not provide a reliable source for the purpose of territorial monitoring.

Figure 10 shows an overall trend of light dispersion in small—and medium-sized cities, while large cities did not change their compactness index over time. Similarly, Figure 11 shows that the urban population density analysis results show a strongly decreasing trend, but this time, it is for all city sizes. The two measures indicate an overall trend of dispersion, mostly among medium-sized cities. Figure 12 presents the outcomes of shrinkage/growth score for each city size type.

4.2. Correlation Analysis

This section presents the outcomes of the correlation analysis between the score and spatial measures such as compactness index (Ci) and urban population density (Ud). They are additionally broken down into more detailed analyses looking at these correlations for urban municipalities (6, 1) and urban–rural municipalities (6, 3). This was done because the characteristics of a single urban area differ from those of an urban-rural area, which considers a more diverse context with possible smaller settlements included in the calculation.

4.2.1. Correlation between the Shrinkage/Growth Score and Compactness Index

An examination was performed on the relation between compactness index and urban shrinkage and growth score for three periods. All statistically significant results are shown in greyscale in the tables. The results for all medium-sized cities changed over time from a minor positive in the first period to a slight negative correlation in the later periods. However, with a p-value above 0.05, all the results are statistically insignificant (

Table 7. Correlation outcomes between the shrinkage and growth score and urban compactness for all medium-sized cities.

	Periods
	2006–2011	2011–2016	2016–2021
Number of observations	180	184	180
Correlation for “urban fabric”	0.09 (p = 0.21)	−0.07 (p = 0.32)	−0.06 (p = 0.4)
Correlation for “artificial surfaces”	0.05 (p = 0.44)	−0.08 (p = 0.235)	−0.08 (p = 0.252)

and Figure 13). When correlated separately between urban municipalities (6, 1) and the score index, there is a minor positive correlation with a strong statistical significance for all periods (

Table 8. Correlation outcomes between the shrinkage and growth score and urban compactness for all medium-sized urban municipalities (6, 1).

	Periods
	2006–2011	2011–2016	2016–2021
Number of observations	147	149	147
Correlation for “urban fabric”	0.34 (p = 0.00)	0.23 (p = 0.00)	0.22 (p = 0.005)
Correlation for “artificial surfaces”	0.29 (p = 0.0002)	0.19 (p = 0.014)	0.18 (p = 0.024)

). This positive correlation means that growing cities are typically more compact. While the strongest correlation was in the first analysed period, this relation weakened over time. The correlation analysis for urban–rural municipalities (6, 3) shows a change from a negative to a positive value. However, the p-value is insignificant for all the periods (

Table 9. Correlation outcomes between the shrinkage and growth score and urban compactness for all medium-sized urban–rural municipalities (6, 3).

	Periods
	2006	2012	2018
Number of observations	33	35	33
Correlation for “urban fabric”	0.02 (p = 0.9)	−0.12 (p = 0.47)	0.02 (p = 0.89)
Correlation for “artificial surfaces”	−0.017 (p = 0.923)	−0.074 (p = 0.67)	0.064 (p = 0.72)

), and, per the criteria described by Onwuegbuzie and Daniel [67], a correlation value dropping from 0.3 to 0.18 is not strong.

4.2.2. Correlation between the Shrinkage/Growth Score and Urban Population Density

The correlation between urban population density (an independent variable) and urban shrinkage and growth score (a dependent variable) was conducted for three periods. The results for all medium-sized cities for the first two periods indicated a slight negative correlation with a statistically significant p-value (

Table 10. Correlation outcomes between the shrinkage and growth score and urban density for all medium-sized cities.

	Periods
	2006–2011	2011–2016	2016–2021
Number of observations	180	184	180
Correlation for “urban fabric”	−0.29 (p = 0.000)	−0.27 (p = 0.000)	−0.20 (p = 0.005)
Correlation for “artificial surfaces”	−0.222 (p = 0.002)	−0.231 (p = 0.001)	−0.105 (p = 0.157)

and Figure 13). However, the results from the last period for “artificial surfaces” show that this correlation disappears. While in the first analysed periods, a lower score (high shrinkage) is related to higher urban density, this trend does not hold, and in 2016–2021, it is not true anymore. Moreover, when looking at the separate correlations conducted for the urban municipalities (6, 1), there is a switch to a positive correlation trend in the last period (

Table 11. Correlation outcomes between the shrinkage and growth score and urban density for medium-sized urban municipalities (6, 1).

	Periods
	2006–2011	2011–2016	2016–2021
Number of observations	147	149	147
Correlation for “urban fabric”	−0.18 (p = 0.022)	−0.11 (p = 0.15)	−0.03 (p = 0.71)
Correlation for “artificial surfaces”	−0.081 (p = 0.32)	−0.016 (p = 0.83)	0.12 (p = 0.14)

). While due to high p-values, these correlations are statistically insignificant. The analysis for urban municipalities in the last period confirms a change in trend. For urban–rural municipalities, it also shows weaker values (

Table 12. Correlation outcomes between the shrinkage and growth score and urban density for medium-sized urban–rural municipalities (6, 3).

	Periods
	2006–2011	2011–2016	2016–2021
Number of observations	33	35	33
Correlation for “urban fabric”	−0.24 (p = 0.16)	−0.26 (p = 0.11)	−0.18 (p = 0.31)
Correlation for “artificial surfaces”	−0.242 (p = 0.17)	−0.28 (p = 0.103)	−0.073 (p = 0.68)

). Nevertheless, similarly to compactness, a correlation value of −0.2 is considered a weak relation. As the trend changes over the analysed periods, it signals a possible positive correlation in the future. As urban density increases, the shrinkage score increases, meaning that higher densities appear to be beginning to be associated with less shrinkage.

5. Findings and Discussion

The decline in Poland’s urban population presents a significant challenge to the country’s spatial policy. The trend of urban shrinkage, understood in a multicriteria way as a socio-economic and demographic decline, is most noticeable in medium-sized cities. However, while some cities struggle to handle the process, others thrive. By understanding the patterns and interconnections between urban form and shrinkage, it is possible to gain insights into the spatial characteristics that influence this phenomenon. This can lead to informed management of the negative impacts and enable sustainable development. In this section of the paper, we seek to check three research hypotheses on relationships between medium-sized city urban form and shrinkage (

Table 13. Summary of hypothesis confirmation.

Hypothesis	Selected Urban Areas (CLC Classes)	All Medium-Sized Municipalities (6, 1 and 6, 3)	Medium-Sized Urban Municipalities (6, 1)	Medium-Sized Urban-Rural Municipalities (6, 3)
H1: There is a statistically significant correlation between urban compactness and shrinkage.	“Urban fabric”	NO	YES	NO
	“Artificial surfaces”	NO	YES	NO
H2: There is a statistically significant correlation between urban population density and shrinkage.	“Urban fabric”	YES	Yes, for 2006–2011	NO
	“Artificial surfaces”	YES for 2006–2016	NO	NO
H3: The trend persists within the analysed timeframe.	“Urban fabric”	YES	YES (for compactness relationship only)	NO
	“Artificial surfaces”	NO	YES (for compactness relationship only)	NO

):

• Addressing H1. There is a statistically significant correlation between compactness and shrinkage of medium-sized Polish cities.
• Addressing H2. There is a statistically significant correlation between urban population density and shrinkage of medium-sized Polish cities.
• Addressing H3. The trend persists within the analysed timeframe.

5.1. Urban Shape Compactness and Urban Shrinkage

The empirical analysis conducted on the urbanised areas in Polish municipalities indicates a discernible declining trend in the compactness index (Ci) over time. Notably, the reduction in compactness is more pronounced in small- and medium-sized cities. Upon considering all types of medium-sized cities (6, 1 and 6, 3), Hypothesis 1 (H1) is not true as no statistically significant correlation was observed between urban-shape compactness and shrinkage. However, this hypothesis only holds true in urban municipalities (6,1), where a statistically significant, positive, moderate correlation exists between urban compactness and the shrinkage/growth score. The findings suggest that a more compact urban municipality is less likely to experience shrinkage. This relationship persists throughout the analysed period from 2006 to 2021 and is attributable to various factors.

A comprehensive understanding of this phenomenon necessitates the examination of diverse urban, economic, social, and demographic dynamics and their influence on sustainable development. Compactness is intricately linked to heightened resilience and sustainability, rendering it beneficial for medium-sized cities with limited governance capacities. The sprawl of urban form and expansion patterns escalate the operational costs of infrastructures. Conversely, compact cities adeptly navigate challenging socio-economic trends by concentrating activities within a smaller area, thereby reducing costs and upholding clear urban–rural boundaries. Consequently, it is observed that the more compact the urbanised area of a city, the lower the likelihood of shrinkage.

Research by Wang et al. [4] underscores that the compactness of shrinking cities in Northeast China is notably lower than that of growing ones. Conversely, a global-scale study by Angel et al. [45] asserts that city-shape compactness is independent of city population, area, population density, and per-capita income but may be contingent on physical barriers, merging of adjacent settlements, inter-city roads and rail lines, land-use restrictions, beachfront preferences, and land-market distortions. Furthermore, Dong et al. [68] suggest that transitioning from a compact form to a dispersed one is common, while the reverse is rare and difficult to achieve. Therefore, prioritising the maintenance of existing urban compactness is imperative for cities contending with urban shrinkage processes.

5.2. Urban Population Density and Urban Shrinkage

The collected data indicate a consistent decrease in urban population densities across all urban municipalities over the observed period, paralleling the findings on urban compactness. Correlation analysis between the density and the shrinkage/growth score among medium-sized cities reveals that denser urban areas were more inclined to experience shrinkage between 2006 and 2016 than less dense ones. Although the correlation results exhibit a considerable scatter, they remain statistically significant throughout that period, suggesting a trend that diminished and eventually vanished in 2016–2021. Hence, Hypothesis 2 (H2) holds true when considering all medium-sized cities, whereby denser urban areas during 2006–2016 demonstrated higher degrees of shrinkage.

This pattern aligns with the broader global trends investigated by Angel et al. [45], demonstrating an overall decline in urban density between 1990 and 2014, particularly prominent in developed countries and associated with higher economic growth rates. While Schwarz et al. [15] linked low-density settlements with the urban landscape of shrinking cities, our study of medium-sized Polish cities does not fully support this view. Nonetheless, it suggests that future development in medium-sized cities might adhere to a pattern described as “growth sprawl” by Siedentop and Fina [25], characterised by initial growth, followed by an outflux of residents from the city centre to the outskirts and subsequent urban perforation and decreased densities. However, the latest analysed period indicates a departure from this trend, as low urban population density no longer correlates with higher shrinkage/growth scores, hinting at the potential emergence of “shrinkage sprawl” [25] with a robust statistical relationship between decreasing density and increasing shrinkage in cities in the near future.

In the context of Poland, the emigration of inhabitants from denser, medium-sized cities to larger metropolitan areas may be attributed to metropolization trends driven by limited job prospects and overcrowded housing conditions. Notably, Poland exhibits consistently high overcrowding rates, a key dimension in evaluating housing conditions [69]. This trend highlights one of the challenges posed by dense urban areas and contributes to migration decisions to access better living conditions at a reduced cost. Large cities in Poland witness substantial new housing construction, attracting increased investment and employment opportunities. These dynamics, coupled with economic liberalisation and land-use policies allowing for suburban sprawl and market-led growth, have resulted in declining densities in medium-sized urban centres from 2006 to 2021 and escalating urban growth rates within the inner areas and suburbs of major cities.

5.3. Changes in Time

The third hypothesis (H3) is valid for the first relationship between shrinkage and compactness and only for urban municipalities. It is also valid for the relationship between shrinkage and density for all medium-sized municipalities when measuring urban density using “urban fabric” as the denominator. When considering “artificial surfaces” for the urban-density calculation, the trend did not persist over all analysed periods but disappeared in the last.

Furthermore, the differences in urban density between the two types of municipalities substantially impact the correlation trend. The heterogeneous nature of medium-sized cities is emphasised, indicating variations in their functioning within the urban network system. These differences are crucial considerations when devising planning implications. According to Xu et al. [70], historical development patterns of cities play a significant role in their current density, with cities historically developed as dense urban centres likely to retain this characteristic even as they experience shrinkage. An example of such a city is Bartoszyce in Northern Poland, which exhibits dense and compact urban characteristics despite facing challenges in attracting capital and residents due to its location (Figure 14). Conversely, newer growing cities, particularly those near metropolitan areas, tend to develop in a sprawl-oriented, low-density pattern by merging with adjacent settlements, as described by Angel et al. [45]. Wieliczka, for instance, exemplifies this pattern, which is characterised by sprawling development in diverse topography. While the city centre retains a relatively dense, historic urban fabric, the predominant development consists of low-density urban sprawl patterns. The location, particularly its proximity to the metropolitan area of Krakow, serves as a significant driver of growth for Wieliczka, offering ample job opportunities. These findings align with the classification of Chinese cities by Liu et al. [22]. They identified five main types depending on the relationship between the hierarchy of cities in the network. Wieliczka exhibits a so-called “growth-driven” model, with Krakow being the growth pole for surrounding medium-sized and small cities. Consequently, these findings highlight the diverse nature of medium-sized cities and emphasise the need to consider their historical development and urban characteristics in urban planning and policymaking.

5.4. Planning Implications

Studies conducted by Schiller and Kenworthy [28] stress the importance of urban densities in creating sustainable cities oriented to public transport, walking, and cycling. Moreover, according to Lytyński’s [29] research, a preponderance of low densities is accompanied by lower GDP in selected municipalities. In shrinking cities, lower housing demand and lower housing costs might be seen as an opportunity to contain the sprawl and focus on improving inner city areas. This can, in the future, attract residents to central areas, maintaining or increasing density. Planning strategies should stimulate it further through investment in public transport infrastructure, better conditions for walking and cycling and improvement of urban life, public spaces, green areas, and social infrastructure [71]. Reurbanisation policies should focus on brownfields, industrial areas, former car parks, and excessive road infrastructure rather than open green spaces.

Newman et al. [72] provide a detailed discussion of the different urban fabrics that exist in all cities today and portray these as walking fabrics, public transport fabrics, and automobile-based fabrics. They show how, over decades, traditional urban fabrics developed in the walking era (up to about 1850) and public transport fabrics (dominant in the period from 1850 to around the Second World War), have been negatively impacted and even destroyed by the imposition of automobile-based urban development, which stresses space-consuming parking, expansive road networks, and single-family, sprawling housing. Their analysis shows the critical need to stop and reverse this process by re-urbanising traditional walking and public transport fabrics, especially by reclaiming roads and parking spaces for public life and new residential development and developing TOD around rail stations. In parallel, they also emphasise the need to minimise further automobile-based urban development in most parts of the city.

Through such approaches, well-managed shrinking cities could re-grow or enter a stable and sustainable degrowth path with a balanced economic base and fewer expenses lost on excess infrastructure and excessive travel costs in cars. In essence, they have the potential to become more sustainable and resilient [73,74]. By containing its boundaries and avoiding destructive city sprawl into forested areas, thus reducing the urban heat island effect [71], an urban area can maintain lively urban life, manage assets efficiently, and allocate funds to enhance existing potential rather than further expanding built-up areas. It is worth adding that uneven growth and shrinkage patterns within one city are also common [19,75]. Therefore, it is essential that the local government monitors the situation and identifies problems early.

5.5. Research Limitations

Overall, it is essential to note that the factors mentioned above can interact in complex ways, and the specific reasons for the observed trends in this large sample of medium-sized Polish cities would require detailed local-scale analysis and additional qualitative studies. Data-collection methods and definitions for measuring “urban compactness” and “urban density” can significantly affect the observed correlations. Despite showing two approaches to selecting different CLC land-use classes, the urban area defined in this study can be seen as overly generalised. Therefore, achieving conclusive urban-density results may require a different or refined method, perhaps collecting population and urban land-area data from local statistical databases. However, such an exercise for hundreds of cases would be highly time-consuming and beyond the scope of the present research.

Additionally, there are risks associated with data accuracy, particularly about census data in Poland. For instance, population distribution data in Poland could be improved, as there are discrepancies between actual and registered home addresses, especially among young mobile people who still need to update their addresses in the registry. This could result in overestimating the number of inhabitants in medium-sized cities, as younger population cohorts tend to relocate to larger centres for work, education, and social opportunities unavailable in smaller settlements.

Finally, while the method presented in this paper is general enough to allow for country-wide spatial analysis of all available cases, it may need to be revised to account for specific topographical conditions. For example, natural elements such as hills or bodies of water may define the boundaries of a city and render the compactness index of a city’s shape and urban population density misleading. Therefore, caution should be exercised when applying the method to such cases.

5.6. Implications for Further Research

Although this study did not account for the nuances of cities within specific topographical settings, further research is necessary to explore these relationships through qualitative data. The discrepancies between the urban population densities identified in this article should be further investigated.

Additionally, a more nuanced understanding of compactness can be achieved by examining the compactness of selected cities through a functional lens and utilising data on activities such as the Internet of Things (IoT) and nighttime patterns. Further, analysing the socio-economic and cultural attributes of shrinking cities can provide a more holistic comprehension of which areas of the city face difficulties and the underlying causes.

Moreover, future studies could focus on morphological and urban qualities, such as building density, typology, floor-area ratio, street connectivity, and expansion pattern, which play a significant role in better understanding an area’s compactness. Future studies could also benefit from a more extensive timeframe and a broader scope with a comparative view juxtaposing two or more national contexts and looking at different city sizes in search of patterns.

6. Conclusions

The present study provides empirical insights into the intricate relationship between urban form and shrinkage in medium-sized Polish cities. It delves into the pressing issue of urban shrinkage, examining the interplay between urban form, compactness, and population density. Spanning 2006 to 2021, our analysis aimed to test three hypotheses to grasp these relationships. The results revealed a nuanced and dynamic relationship between urban form and shrinkage. While no significant correlation was found between urban compactness and shrinkage across all medium-sized cities, a positive correlation was identified in urban municipalities. This suggests that more compact urban areas are less susceptible to shrinkage, underscoring the potential advantages of compact urban forms for resilience and sustainability. Conversely, denser urban areas showed a higher likelihood of shrinkage between 2006 and 2016. However, this trend weakened in the subsequent years, which may signal a reversal of the earlier relationship in medium-sized Polish cities.

The study emphasises the diverse nature of medium-sized cities and the need for tailored urban-planning strategies. Compactness and density, while not consistently linked to shrinkage, play roles that demand careful consideration in sustainable urban development. This is especially so in supporting global efforts to develop more sustainable urban transport by reducing trip distances, making walking, cycling, and public transport more attractive and feasible, and minimising the need to use a car.

Based on Corine Land Cover (CLC) spatial data, our methodology extends to urban areas across Europe, providing a valuable framework for broader spatial analyses. However, the observed correlations warrant further investigation through detailed local-scale analyses and additional qualitative studies involving more diverse variables than urban density and compactness alone. Recognising limitations in data accuracy and methodological constraints, the study urges caution in interpreting observed correlations as causative, highlighting the multifaceted nature of urban-shrinkage determinants. Due to its widely recognised adverse sustainability implications, it also discourages viewing urban sprawl as a remedy to shrinkage.

Future research endeavours can benefit from refining compactness assessments to encompass morphological and functional dimensions and longer time spans, leveraging advancements in urban data analytics. In conclusion, the study underscores the need for further research, particularly in exploring the socio-economic and cultural dimensions of shrinking cities. It calls for nuanced housing policies and land-use regulations to manage urban density in shrinking cities without compromising their long-term development or sustainability objectives.